Experiment example: Kidlang

Author: Sherine Bou Dergham

Description

This experiment is implemented using the Presentation framework.

The study is about analysing brain signals in two states:

  • a context is given

  • no context is given

In each state there are two conditions. The type of used stimuli is auditory and visual.

In the context (also called dialogue) task:

  • a visual stimulus shows you two objects colored differently

  • in the phrase condition, the computer will say both words for object name and color (e.g., Red House) and the participant will say both words for the other object (green cup)

  • in the adjective/object (called noun-only) condition, the computer will say “mmmm”, Red (the computer will hide the name of the object or the adjective), and then the participant should “Green”

  • there is a third condition serving as baseline, where the computer is simply saying “mmmm”

In the non-context given task (listening task):

  • the computer will name an object, and the participant has to select the object that was named from a visual stimuli

  • There are two conditions:

    • computer will say object+adjective and the participant has to select the object (the visual stimuli to pick from are colored objects)

    • the computer will say “mmmm” glass and the participant has to select the object (the visual stimuli to pick from are colorless objects)

We care about the variables “dialogue” and “listening”

Time window of analysis:

  • each trial for dialogue and listening is -0.7 to 2 seconds, where 0 corresponds to ?

In the Presentation code, the trigger ID’s are:

  • Dialogue: - Phrase: 18 - Object: 34

  • Listening: - Phrase: 20 - Object: 36

Need to map the pulses to the event ID from the mne GUI, From the mne kit2fiff GUI, to detect correctly the upper labels, we need to set the following settings: Peak (0 to 5V) and value coding: Little-endian. When you press Find events on the GUI, it computes the STI 014 channel according to the parameters you set in the GUI (trough, peak, little endian etc…) In the pipeline, we constructed the same parameters using the pipelinemne_pipelineskit_general_pipelinespipeline_config_fileskidlang_config_file.yml

The next step is to make sure the produced .fif has a correctly constructed STI 014 channel, in which we have the ID’s above.

Only channels from MISC 001 to MISC 008 are used to code the triggers

21 subjects total data acquisitions

  • 20 subjects have a count of 1000 pulses over all their .con files from all the trigger channels (224-231)

  • 1 subject Y0392 have a count of 946 counts, apparently the recording was stopped for the “reading” task, but should be good for the dialogue/listening task

12 trigger ID’s should be found because (3 conditions for every task and four tasks = 12), but for this study we care about 4 of them only. The acquired data has 50 events per run. The total number of events per file is therefore. Meaning there should be 600 trigger codes per data acquisition session (could span over multiple files if the participant took alot of breaks or completed the task over multiple days).

  • verified that 20 subjects have 600 trigger codes

  • 1 subject Y0392 has less than 600 trigger codes (but should have what’s needed for the tasks that we will analyse)

The stimuli has 6 objects and 6 colors

(ICA before and filter of 50Hz with harmonics and bandpass most probably)

Next is to compute the epochs for the events

Stimulus Code access

Available upon request. 6 different .exp files that use the four same .sce files (each is for dialogue, listening, reading, production). The .sce are saved on Gdrive. The .exp is shuffled for participants, first participant uses the first .exp, second participant uses second .exp etc… For each subject, the order of the four tasks would change.

Data access

Acquired datasets are stored safely on NYU Box under kidlang.

Local data on MEG workstation under: D:MEG_Datakidlang

Analysis results

Operations that are already performed:

Not documented yet.